The future of human agency in an AI‑mediated world

AI is moving from the edges of work into the centre of how decisions are made in law firms, financial institutions, and public bodies. It now drafts documents, screens matters, triages risk, allocates work, and increasingly suggests what should happen next. The uncomfortable question for boards is simple: when a system recommends, predicts, or triages, whose decision is it really?

In this essay, “human agency” means our ability to notice what is going on, weigh options, and choose a course of action we are prepared to own—ethically, professionally, and legally. The real risk is not a science‑fiction takeover; it is a slow erosion of that ownership as more of the heavy lifting is handed to opaque tools.

Through the lens of Resonance, the aim is not to resist AI but to design a partnership where people stay clearly in charge of the hard calls while using machines to extend their field of view. In other words, AI should widen our awareness and sharpen our judgment, not dull it.

1. From “human vs machine” to augmentation

Most public debate still asks: will AI replace lawyers, compliance officers, or risk leaders? It is a blunt, zero‑sum frame that belongs to the industrial age. For boards, the more useful question is: where should AI sit in each decision, and where must a human stay firmly on the hook?

A more productive lens is the augmentation paradigm. In plain terms, AI is at its best when it extends human capability - helping us see more data, spot patterns faster, and consider options we would otherwise miss - while leaving the final judgment with a named person.

You already live with this extended‑mind reality. Phones, search engines, document management systems, and precedent libraries are all cognitive extensions. Work on the “extended mind” in philosophy and cognitive science simply makes this explicit: humans naturally offload parts of thinking into their environment, from notebooks to navigation apps. That does not mean we lose our minds to the tool; it means the tool becomes part of how we think.

The difference with AI is not that it extends us, but how far and how opaquely it extends us. Tools that simply surface more information still leave the user clearly in charge. Systems that start to filter options, rank people, or recommend sanctions without clear explanation start to shape the decision itself. At that point, it is no longer enough to say “the partner signed it” or “the officer approved it” if you have not explicitly designed the human‑machine split.

2. Resonance, Consonance, and Dissonance in AI‑mediated decisions

The Resonance model gives a simple way to think about AI adoption without drifting into either hype or fear.

In a Resonant system, human judgment, AI tools, and organisational architecture are tuned to help one another. People understand what a system can and cannot see, where it is used, and who remains accountable for the outcome. Metrics support learning, not blame, and dissent about the tool’s behaviour is welcomed as data.

In a Dissonant system, there is a growing gap between what the organisation says about AI and how it actually uses it. Policies talk about responsibility and ethics, but in practice no one can explain how a model works, who can challenge it, or who owns a bad decision shaped by it. That say‑do gap is classic Dissonance.

In a brittle Consonance, everything appears calm because nobody wants to ask awkward questions. Dashboards look impressive, AI projects are celebrated at town halls, and concerns about bias, explainability, or accountability are quietly sidelined. The apparent harmony rests on silence, not genuine alignment.

For COLPs, COFAs, and board members, the practical task is to move AI programmes out of Dissonance and brittle Consonance into a more Resonant state. That does not require a PhD in machine learning; it requires a clear view of where decisions are being shaped, which Beacons they touch (authority, psychological safety, measurement, accountability), and whether human agency is being amplified or hollowed out.

3. When AI enhances agency – and when it undermines it

Not all AI systems have the same impact on human agency. The key distinction is simple:

• Enhancement happens when AI expands the capacity for informed, reflective choice by providing better information, revealing patterns, or widening the set of options we can reasonably consider.
• Undermining happens when AI quietly narrows awareness, nudges behaviour in ways people do not understand, or creates dependencies that make it hard to exercise independent judgment without the tool.

Examples of enhancement in a legal or regulatory context include:

• Document review tools that surface relevant clauses faster, but require the lawyer to interpret them.
• Risk engines that flag unusual matter profiles or clients, while leaving the decision to escalate with a named person.
• Analytics that show trends in complaints, claims, or regulatory breaches, providing a better map for human decision‑makers.

Examples of undermining include:

• Scoring models that rank clients, staff, or cases without meaningful explanation, where decision‑makers simply follow the “score”.
• Automated triage systems that route matters, allocate work, or set sanctions with no clear override path and no audit trail of human intervention.
• “Assistant” tools that draft advice or key communications where, in practice, time pressure turns review into rubber‑stamping.

The preservation of human agency in an AI‑mediated world does not require the absence of technological influence. It requires conscious, transparent participation in deciding how that influence operates. Agency is not a feeling; it is an architectural choice.

4. The accountability problem: algorithmic abdication vs aligned authority

One of the most dangerous cultural shifts with AI is not technical at all. It is the temptation toward algorithmic abdication of responsibility - the quiet belief that if “the system” said so, the human can stand further back. In Resonance terms, this is a new expression of “Accountability is a Threat to be Diffused”.

In a Dissonant organisation, accountability is already pushed downward or outward: into complex structures, third parties, or faceless committees. AI can become just another layer in that chain—an extra step between the decision and the person who might otherwise own it.

A Resonant approach does the opposite. It treats AI as an artefact that must sit inside an Aligned Authority & Accountability Beacon:

• For each high‑stakes use of AI, there is a clearly named human steward.
• That steward understands the tool’s purpose, limits, and failure modes in broad terms.
• The system is designed to keep a human in the loop where consequences are serious or ambiguous, with a genuine ability to override and a duty to question.

“Human‑in‑the‑loop” is more than a slogan. It is a governance design pattern borrowed from safety‑critical domains like aviation: automation can assist, but in ambiguous, high‑stakes scenarios, the human must retain the final, non‑delegable authority. The same logic applies to AI in professional judgment: if the outcome carries ethical, legal, or life‑changing weight, a person needs to be clearly on the hook.

5. Practical questions for boards and risk owners

Instead of abstract debates about “good” or “bad” AI, boards can make progress by asking a small set of plain‑English questions wherever AI touches meaningful decisions:

• Where is AI already shaping decisions? List the workflows where tools recommend, rank, route, or triage. Include “soft” influence (suggested wording, default options, scores) as well as hard decisions.
• For each of those, who is the named human owner of the outcome? Not a committee or a job title in theory, but a steward who understands they are responsible for how the tool is used and how its outputs are applied.
• What can the system see – and what can’t it see – that a human should care about? Data is always a partial view. A Resonant design makes those blind spots explicit so humans can bring context, ethics, and lived experience back into the frame.
• How easy is it, culturally and technically, to challenge or override the tool? If people feel they will be punished for questioning the system, or if overriding it is practically impossible, you have created Dissonance: a gap between stated values and lived reality.
• Could you explain this chain of responsibility to a regulator or a court? If you had to stand behind a decision heavily shaped by AI, could you show who owned it, how the system was meant to be used, what checks were in place, and where human judgment entered?

These questions connect AI governance directly to existing duties around accountability, ethics, and client care. They also expose where the organisation is drifting into mechanistic thinking—treating people as operators of a machine rather than stewards of judgment.

6. Designing for Resonance: AI as a partner in judgment

The Philosophy of Resonance offers a simple test for AI projects: does this tool make it easier for our people to notice what matters, talk about it openly, and act in line with our stated principles—or does it pull in the opposite direction?

Resonant AI design has a few recurring features:

• Transparency at the right level. You do not need to expose every line of code, but you do need to explain what the system is for, what data it uses, where it is strong, and where it is weak.
• Human‑centred authority. High‑stakes calls have clearly identified human decision‑makers. The system supports them; it does not replace them.
• Feedback loops. Users can report surprising, harmful, or biased behaviour, and the organisation actually responds. This turns frontline experience into learning, not noise.
• Psychological safety. People are encouraged - and protected - when they challenge the output of a model. Dissent is treated as data, not disloyalty.

In that frame, AI is not a rival intelligence but part of the system’s plasticity: its ability to learn, adapt, and reshape itself in response to changing reality. Neuroplasticity (human learning), digiplasticity (machine learning), and organoplasticity (the organisation’s ability to re‑shape itself) can either work against each other or in concert. Resonance is what it looks like when they amplify one another.

Conclusion: agency as a design choice, not a casualty

The future of human agency will not be settled by the raw power of AI models. It will be decided by the mundane but decisive choices leaders make about where authority sits, how transparent systems are, and how much courage they have to keep humans visibly at the centre of high‑stakes decisions.

If AI is introduced as an unquestionable oracle, people will slowly stop exercising judgment, and responsibility will become impossible to locate when things go wrong. If AI is introduced as a powerful but bounded partner - one that extends human sightlines while leaving humans on the hook for the call—then agency is not a casualty of progress but one of its main beneficiaries.

Designing for Resonance means treating AI as part of the social architecture of the firm, not just a new tool. It means asking, in every deployment: does this help our people see more, decide better, and own the outcome - or does it pull them further away from the decisions made in their name? The answer to that question, not the latest technical breakthrough, will shape the real future of human agency.